additional slide for learning rate

[JuliaHerbinger]

add a slide at the end of "gradient boosting concept" chunk about line search vs. constant learning rate and explain how line search works and why we use a constant learning rate.

[lisa-wm]

add from issue 896:
We always treat the learning rate as individual per iteration which is calculated by line search. But I think most (maybe all) implementations use a constant learning rate. Maybe it's better to introduce everything with a constant learning rate and mention in the end that the learning rate can also be calculated by line search.